<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Bruno Mentges — Writing</title><description>Field notes on data engineering and AI by Bruno Mentges — senior engineer writing about production systems, LLM integration, and platform design.</description><link>https://bmentges.github.io/</link><language>en-us</language><item><title>The Knowledge Base Strategy: Giving AI Agents a Memory That Compounds</title><link>https://bmentges.github.io/posts/knowledge-base-as-agentic-memory/</link><guid isPermaLink="true">https://bmentges.github.io/posts/knowledge-base-as-agentic-memory/</guid><description>Why AI coding agents produce dramatically better development plans when they have a structured wiki about your architecture, decisions, and concepts — and an open-source template to build one yourself.</description><pubDate>Fri, 17 Apr 2026 00:00:00 GMT</pubDate><category>ai-agents</category><category>knowledge-base</category><category>second-brain</category><category>open-source</category></item><item><title>I Integrated AI Agents Into Our Data Pipelines. Here&apos;s What Actually Worked.</title><link>https://bmentges.github.io/posts/ai-agents-in-data-pipelines/</link><guid isPermaLink="true">https://bmentges.github.io/posts/ai-agents-in-data-pipelines/</guid><description>The agents were good. But they hit a ceiling I didn&apos;t expect — and the thing that broke through it wasn&apos;t a better model. It was a knowledge base.</description><pubDate>Thu, 16 Apr 2026 00:00:00 GMT</pubDate><category>ai-agents</category><category>data-engineering</category><category>mcp</category><category>production</category></item><item><title>MCP, Agentic Workflows, and Guardrails: A Production Field Guide</title><link>https://bmentges.github.io/posts/mcp-agentic-workflows-guardrails/</link><guid isPermaLink="true">https://bmentges.github.io/posts/mcp-agentic-workflows-guardrails/</guid><description>What Model Context Protocol actually looks like in a real engineering workflow — how to set up context servers, build guardrails that earn trust, and integrate agents where they matter.</description><pubDate>Wed, 15 Apr 2026 00:00:00 GMT</pubDate><category>mcp</category><category>ai-agents</category><category>guardrails</category><category>production</category></item><item><title>The Medallion Architecture Is Not What You Think It Is</title><link>https://bmentges.github.io/posts/medallion-architecture-misunderstood/</link><guid isPermaLink="true">https://bmentges.github.io/posts/medallion-architecture-misunderstood/</guid><description>Everyone talks about bronze/silver/gold layers. Most implementations miss the point entirely. Here&apos;s what the pattern is actually for — and when to break the rules.</description><pubDate>Tue, 14 Apr 2026 00:00:00 GMT</pubDate><category>medallion</category><category>data-architecture</category><category>dbt</category><category>warehouse</category></item><item><title>Building a Data Platform From Zero: A Field Guide</title><link>https://bmentges.github.io/posts/building-data-platform-from-zero/</link><guid isPermaLink="true">https://bmentges.github.io/posts/building-data-platform-from-zero/</guid><description>What I learned designing and building a complete data platform for a growing company — from first table to production analytics. The decisions that matter, the order that works, and the mistakes everyone makes.</description><pubDate>Mon, 13 Apr 2026 00:00:00 GMT</pubDate><category>data-platform</category><category>data-architecture</category><category>bigquery</category><category>greenfield</category></item><item><title>How I Reduced a Petabyte Pipeline&apos;s Cost by 40%</title><link>https://bmentges.github.io/posts/petabyte-pipeline-cost-optimization/</link><guid isPermaLink="true">https://bmentges.github.io/posts/petabyte-pipeline-cost-optimization/</guid><description>Performance engineering at scale isn&apos;t about clever tricks — it&apos;s about understanding economics. A case study in Spark optimization, partition strategy, and thinking in dollars instead of milliseconds.</description><pubDate>Sun, 12 Apr 2026 00:00:00 GMT</pubDate><category>spark</category><category>optimization</category><category>cost</category><category>petabyte-scale</category></item><item><title>The 70% Bug Squash: What One Hackathon Taught Me About Tech Debt</title><link>https://bmentges.github.io/posts/hackathon-tech-debt-bug-squash/</link><guid isPermaLink="true">https://bmentges.github.io/posts/hackathon-tech-debt-bug-squash/</guid><description>How a 2-day hackathon at a Brazilian e-commerce company squashed 70% of outstanding bugs and improved uptime from 92% to 99%. The real lesson isn&apos;t about hackathons — it&apos;s about permission.</description><pubDate>Sat, 11 Apr 2026 00:00:00 GMT</pubDate><category>tech-debt</category><category>hackathon</category><category>engineering-culture</category><category>leadership</category></item><item><title>From 1.5 Billion Records/Day to AI Agents: My 28-Year Engineering Arc</title><link>https://bmentges.github.io/posts/28-year-engineering-arc/</link><guid isPermaLink="true">https://bmentges.github.io/posts/28-year-engineering-arc/</guid><description>From telecoms in 1998 to AI-augmented data engineering in 2026. What stayed the same, what changed completely, and what I wish I&apos;d known earlier.</description><pubDate>Fri, 10 Apr 2026 00:00:00 GMT</pubDate><category>career</category><category>retrospective</category><category>data-engineering</category><category>ai-engineering</category></item></channel></rss>